Humane Ingenuity 32: Faint and Loud Signals
by Dan Cohen
If for some reason you could use some relaxation right now, I recommend heading over to Faint Signals, an interactive work of art that was one of the clever entries in the annual competition run by the British Library Labs for creative reuses of their collection.
Faint Signals generates an imagined Yorkshire forest, which you can then explore through the seasons. As you meander through the digital woods, peaceful natural sounds from the British Library's extensive audio collection—birds, rain, wind—are encountered. Faint Signals doesn't exactly rival the real signals of the real thing, but HIers, it's winter, and there's a pandemic still going on. So put your headphones on, turn your phone off, and take a leisurely stroll through the virtual forest.
OpenAI has released CLIP, the Contrastive Language–Image Pre-training tool, which connects some of their work on natural language supervision (see prior HIs on GPT-2 and GPT-3) with image analysis, forging associations between the textual and the visual.
Travis Hoppe used CLIP on some famous poems and an open access collection of landscape photos. The results are often uncanny, finding strikingly appropriate nature photographs for each line of poetry:
Where Alph, the sacred river, ran through caverns measureless to man Down to a sunless sea.
And on the pedestal, these words appear: My name is Ozymandias, King of Kings.
’Twas brillig, and the slithy toves Did gyre and gimble in the wabe: All mimsy were the borogoves, And the mome raths outgrabe.
Because I could not stop for Death – He kindly stopped for me – The Carriage held but just Ourselves – And Immortality.
Melissa Terras and David Beavan created Windsor-o-Tron, which can generate (using GPT-2 and and a data set of old speeches) as many Christmas addresses as the Queen might need:
Our lives are shaped by our past, and as we live out our future together we should know each other best. It is difficult for us to know far into the future as our families gather round us, but it is better that we have some sense than that we have any sense at all. I wish you all, together with your children and grandchildren, a blessed Christmas.
(Also of note: Europeana has a call out for the assembly of new cultural heritage data sets for AI tools to train on. Keeping on eye on that.)
Readers of Humane Ingenuity know that I care a lot about the preservation of, and access to, a wide array of human expression. From my notepad, here is a bit of my running list of what we will have, and have not, from the past few weeks:
- We will have digital facsimiles of books, and pure, reflowing ebooks, from works that were published in 1925, but potentially not from works that were published in 2020, because of copyright law, digital rights management/encryption, and the ebook market that libraries currently face, in which they often rent rather than buy.
- We will have very few literary works written in Flash, which reached the end of its technical life in 2020, and those works of electronic literature that do survive will be preserved by a select few.
- We may have the contents of an entire social network frequented by a large number of unethical people, but only because a hacker took action before it was offlined by Amazon Web Services. And if it survives, it will be in the dark corners of hidden storage systems, not in a preservation institution and widely available to future researchers. (This may be ok with many of you, and maybe with me too; my notepad makes no judgments.)
- We will likely lose access to most of a much larger and older social network, because the official archiving of that network ended on December 31, 2017. Ethical archivists have found ways to save the IDs for messages from that network since 2017, but these "dehydrated" formats, while laudably giving agency to the creators of those messages rather than just rashly grabbing everything, also makes future retrieval ("rehydration") iffy. (This may also be ok; the notepad has no comment.)